1. Field of the Invention
The present invention relates to a network management system and a network management method capable of controlling an agent even when a fault occurred on a logical communication channel between a manager and the agent.
2. Description of the Related Art
The conventional network management system of this kind forms a hierarchical network management system provided with a sub-manager between an agent (management function module of a managed node) and a manager (management function module of a management node, and hereinafter referred to as an integration manager) in a large-scaled network management by use of an (SNMP). Management objects in the management range of a sub-manager are periodically collected under the control of the sub-manager, through the mediation of each agent belonging to the management range, and this collected information is notified to an integration manager in reply to a reference request from the integration manager, thereby controlling the management objects of each agent.
This kind of the conventional technique is disclosed in, for example, Japanese Patent Publication Laid-Open (Kokai) No. Heisei 7-334445, "Hierarchical Network Management System". Logical relationship between agents, sub-managers, and an integration manager in a network management system disclosed in the same publication is shown in FIG. 12. In the management system shown in FIG. 12, between a sub-manager 1210a and each agent 1220a, 1221a, the management objects are controlled by use of the SNMP and ICMP (Internet Control Message Protocol) that are management standards of the IAB (Internet Activities Board). Between the sub-manager 1210a and an agent non-installed IP node 1230a, the management objects are controlled by use of the ICMP. The sub-manager 1210a is connected with a collected MIB database 1240a holding a plurality of management object groups collected through the agents in the management range in the MIB (Management Information Base) format of tree structure. A sub-manager 1210b, an agent 1220b, an agent non-installed IP node 1230b, and a collected MIB database 1240b are connected with each other in the same logical relationship. Each sub-manager is connected to an integration manager 1200.
In a network management system constituted as mentioned above, the respective sub-managers 1210a and 1210b are provided with the SNMP agent function toward the integration manager 1200 and the information of the collected MIB database is notified in reply to a reference request from the integration manager.
The above-mentioned conventional network management system, however, sets up one unit as a manager (sub-manager) to which each agent notifies the management objects. Therefore, when a fault has occurred on a logical communication channel between a sub-manager and an agent, the agent cannot be controlled by the sub-manager and an integration manager disadvantageously, in spite of actually being in the state capable of notifying the management objects. The similar disadvantage may occur between a sub-manager and an integration manager.
Since a small-sized network containing the small number of agents needs no sub-manager originally prepared for dispersing the processing load of a manager, a hierarchical network management system consisting of an integration manager and sub-mangers would not necessarily improve the quality of the network management.
Further, in case of a hierarchical network management system consisting of an integration manager and sub-managers, a logical communication channel between an integration manager and a sub-manger and a logical communication channel between a sub-manager and an agent must be both in the normal condition in order that the integration manager might control the agents. Therefore, in a small-sized network naturally requiring no sub-manager, the quality of the network management would deteriorate on the contrary.